sequify-server

Sequify

Add AI assistants to your website or app with just a few lines of code.

中文文档 · Documentation · Examples

What is Sequify?

Sequify is an AI assistant SDK that lets your website or app users control the interface with natural language. Instead of AI "looking at" your page (screenshots, DOM parsing, guessing coordinates), your website tells AI what it can do, and AI calls those capabilities directly.

The result: Deterministic execution, minimal token usage, blazing fast response times.

Traditional: AI → sees page → guesses → clicks → sees result → guesses again
Sequify:     Website registers capabilities → AI calls functions → precise execution

✨ Key Features

🏗️ Architecture

graph TB
    subgraph "User Browser"
        Page[Web Page]
        SDK[Frontend SDK]
        Page -->|Register capabilities| SDK
    end

    subgraph "Developer Backend"
        Proxy[Server SDK<br/>Auth + Proxy]
    end

    subgraph "Sequify Server - Docker"
        WS[WebSocket Server]
        Agent[Eino Agent<br/>ReAct Loop]
        Queue[Task Queue<br/>Asynq]
        WS --- Agent
        Agent --- Queue
    end

    subgraph "Shared Storage"
        DB[(Database<br/>PostgreSQL/MySQL/SQLite)]
        Redis[(Redis<br/>Pub/Sub + Cache)]
    end

    subgraph "LLM Provider"
        LLM[OpenAI / Claude<br/>or any OpenAI-compatible API]
    end

    SDK -->|WebSocket| Proxy
    Proxy -->|WebSocket| WS
    Agent -->|API calls| LLM
    WS --> DB
    WS --> Redis

🚀 Quick Start

Get started in 60 seconds with Docker:

# 1. Start Sequify Server
docker run -d \
  -p 8080:8080 \
  -e SEQUIFY_LLM_PROVIDER=openai \
  -e SEQUIFY_LLM_API_KEY=your-api-key \
  -e SEQUIFY_LLM_MODEL=gpt-4 \
  sequify/server:latest

# 2. Add Server SDK (Go example)
go get github.com/sequify/server-sdk-go

# 3. Add Frontend SDK
npm install @sequify/web-sdk

That's it. Now register capabilities on your frontend and start handling user intents.

Example: Register a Capability

// frontend.ts
import { SequifySDK } from '@sequify/web-sdk';

const sdk = new SequifySDK({
  wsUrl: 'wss://your-server.com/sequify',
});

// Register what your page can do
sdk.register({
  name: 'search_products',
  description: 'Search products by name, category, or price range',
  params: {
    type: 'object',
    properties: {
      query: { type: 'string', description: 'Search query' },
      minPrice: { type: 'number', description: 'Minimum price' },
      maxPrice: { type: 'number', description: 'Maximum price' },
    },
    required: ['query'],
  },
  execute: async (params) => {
    // Your existing search logic
    return await searchProducts(params.query, params.minPrice, params.maxPrice);
  },
});

User says: "Find me the cheapest phone case under $20"

AI calls: search_products({ query: "phone case", maxPrice: 20 })

Result: Precise, deterministic execution. No screenshots, no DOM parsing.

🆚 How It Differs

Sequify is for your own products. Browser automation is for scraping other people's sites.

💼 Use Cases

🛠️ Tech Stack

📚 Documentation

• Product Overview — What Sequify is, architecture, competitive analysis
• Core Flows — Connection lifecycle, capability registration, intent execution
• Server Deployment — Docker deployment, configuration, operations
• Protocol Specification — WebSocket protocol details
• Coding Standards — Development guidelines

🤝 Contributing

We welcome contributions! Please see CONTRIBUTING.md for guidelines.

Current status: Single-developer mode. Full PR review process will be enabled when collaboration begins.

📄 License

MIT License

🌟 Show Your Support

If Sequify helps you, consider giving it a star! It motivates continued development.